PhD position in Computational Semantics (1.0 FTE)

The Utrecht Institute of Linguistics (UiL OTS) invites applications for one PhD position in the area of computational semantics within the research programme Forests and Trees: the Formal Semantics of Collective Categorization. This is a five-year research programme funded by an Advanced Grant from the European Research Council (ERC), and led by Prof. Yoad Winter (Principal Investigator). The project aims to investigate the mechanisms underlying our linguistic ability to conceptualize collections. This use of language is exemplified by the sentence "the Rockies are near" which categorizes a collection of mountains by estimating distance from the nearest mountain in the Rockies. Collective categorization is ubiquitous in language. It forms an important probe into the connections between grammar and the mind, and is crucial for artificial intelligence. For instance, a command to a domestic robot such as "take those books out of their packages and arrange them alphabetically on the shelf" requires the robot to act on individual books while understanding internal relations within the collection. Similarly, spatial collective categorization is also important for natural language interfaces of Geographic Information Systems.

The research programme takes an interdisciplinary approach and makes use of theoretical and empirical insights from linguistics and psychology, as well as formal and computational methods from computational linguistics/natural language processing. The programme consists of three work packages: in psycholinguistics, formal semantics and computational linguistics. This job opening concerns a PhD position within the computational work package, which is carried out with the collaboration of Dr. Tejaswini Deoskar (ILLC/UvA and Utrecht University) and Dr. Joost Zwarts (Utrecht University).

PhD Project - Collective categorization using geometric features of concepts

This project will make use of large language corpora and modern machine learning methods in order to analyse spatially-based concepts expressed in language ("the Rockies are far", "the books are inside the box", "the soldiers are surrounded"). The project's methodology will involve an interaction between data-driven machine learning and semantic parsing approaches, as well as principles in formal semantics for analysing spatially-based collective categorisation. The candidate for this position should have a strong background in artificial intelligence, machine-learning and computer science as well as an interest in formal semantics. The successful applicant will be supervised by Dr. Tejaswini Deoskar and the PI, in collaboration with Dr. Joost Zwarts. The PhD candidate will have the opportunity to study the above topic based on their own expertise and research ideas, while keeping to the broad goals of the project. The project will also involve the opportunity of collaborating with a broader team of NLP researchers, especially from the Language and Computation group at the ILLC, University of Amsterdam (Prof. Khalil Sima'an, Dr. Jelle Zuidema) and from City, University of London (Prof. James Hampton, Psychology).

Tasks for the PhD candidate will include:

completion and defence of a PhD thesis within four years; regular presentation of intermediate research results at workshops and conferences; publication of at least two peer-reviewed articles in established international journals or conference proceedings; help with organisational tasks connected to the project, such as the organisation of conferences and workshops; participation in all training programmes and expert meetings scheduled for the project; participation in selected training programmes scheduled for the Research Institute, Graduate School and the National Research School.